基于多源同步数据的闽江下游悬浮物定量遥感

利用2006-09-18的同步Landsat TM数据、水面实测光谱数据和现场水样数据,研究了闽江下游的悬浮物,在这3种同步数据的基础上建立了分别基于实测光谱与影像光谱的悬浮物遥感预测模型.结果表明,实测光谱数据在690nm波长处,归一化光谱反射率与悬浮物浓度达到最大正相关,由690nm和530nm二处的反射率构成的比值预测模型与实测悬浮物浓度的拟合精度最高,最佳拟合模型可表达为SS=116.2(R690/R530)-33.4.TM影像各波段中以(TM2 TM3)2波段组合与实测悬浮物浓度的相关性最佳.由其所建立的影像光谱预测模型与实测悬浮物浓度的拟合精度最高.最佳拟合模型可表达为SS=3793.7(RTM3 RTM2)2-16.5.模型的精度评价表明,实测光谱模型的预测能力要强于影像光谱模型.但二者差异不大.在缺乏地面实测光谱数据时.基于影像光谱的遥感定量模型仍不失为一种预测悬浮物浓度的有效方法,其对闽江下游悬浮物浓度的反演结果能较准确地反映出该区域悬浮物浓度分布的空间差异,具有较高的实用性.

Quantitative Estimation of Suspended Solid Concentration in the Lower Min River Based on Multi-Source Synchronal Data

Three synchronal data collected on 2006-09-18 have been used in the study of the suspended solid concentration (SSC) of the lower Min River, which are in situ sampled water data, field-spectrometer measured spectral data and Landsat TM spectral data. Two models for predicting SSC have been proposed, one of which is based on field-spectrometer measured data and the other is on Landsat TM data. The statistical analysis of the field-spectrometer measured data has revealed that the reflectance of the SSC at the 690 nm has the strongiest correlation with the in situ-sampled SSC data. The regression model can be expressed as SS=116.2(R690/R530)-33.4. Furthermore, the model built upon the ratio of the reflectance at 690 nm to 530 nm has the best fitness with the in situ sampled SSC data. While the best predicting model for the Landsat TM data is achieved using the band combination of (TM2+TM3)² and is defined as SS=3793.7(R_TM3+R_TM2)²-16.5. The assessment of the two models shows that the model on the field-spectrometer data has higher accuracy than that on the Landsat TM data but the difference is not big. This suggests that the Landsat TM data are still valuable in the prediction of the SSC if the field-spectrometer data are not available. Consequently, the predicting model based on the Landsat data has been applied in the study of the SSC of the lower Min River. The result shows that the model can efficiently reveal the SSC with its spatial distributional pattern features.